COS 333
Back to Project
There are many individuals and groups on or near campus who have really interesting problems that could be profitably attacked by folks in COS 333. Here are some of them. They have come from a variety of sources on campus.
If you're interested in working on one of the proposed projects, then you should contact the proposing organization directly. Typically the organization will agree to work with the first project team that makes contact, so it's important that you make contact soon.
The first proposals are new this semester. They're listed in the order received. To some extent the lead instructor has vetted those projects; but you should make sure that they're right for the course and for you. The others are hold-overs from previous semesters but still of interest. The lead instructor has not vetted those projects.
Pocket Doc is a simple, patient-controlled health record application designed specifically for low-income, vulnerable, and low-health-literacy patients. It would allow patients to bring together medical records from different doctors, hospitals, and health systems, organize the most important information into an easy-to-manage format, identify preventive care they may be due for, and share a concise health history with their clinicians.
The goal is straightforward: give patients who often have the hardest time navigating health care an extraordinarily simple tool for keeping track of their health information and carrying it with them wherever they receive care
Starfield Primary Care is a clinician-owned and operated 501(c)(3) nonprofit primary care organization serving more than 12,000 patients throughout Southern California. We care predominantly for low-income and vulnerable populations, including people experiencing homelessness, medically complex patients, and patients with limited health literacy. Starfield is also in the process of becoming certified as a Federally Qualified Health Center (FQHC).
Patients' medical information is often scattered across hospital systems, physician offices, EHRs, health information exchanges, patient portals, and PDF documents. This fragmentation is especially harmful for patients with fewer resources or limited health literacy, who may be expected to remember complicated histories or navigate multiple portals. Pocket Doc would give these patients one simple place to organize and understand their medical information.
Apple, startups such as Torch, and major EHR systems such as Epic and Oracle Health/Cerner are developing ways for patients to access or aggregate health data. Pocket Doc would be different because it would be designed first for low-income and low-health-literacy patients, independent of any single health system, and operated by a nonprofit clinician-led organization with no interest in selling patient data or using the application as a vehicle for advertising or monetization. Long-term sustainability would rely on grants and philanthropic or donation-based support, potentially resembling the Wikipedia model.
A major advantage for the COS 333 team is access to real-world users. Starfield can provide synthetic and de-identified test data and, with appropriate privacy protections and consent, recruit patients from the population Pocket Doc is intended to serve for usability testing. Students can iteratively test whether the application is truly understandable and usable for people with limited health literacy or technical experience.
Teddy Rountree-Shtulman would serve as the primary liaison between Princeton and Starfield. Michael Hochman, MD, MPH, Princeton Class of 2001 and Starfield’s founder and CEO, would provide clinical and organizational guidance, and Millie Batra, Starfield’s 2026–2027 Clinical Innovation Fellow, would help coordinate product development, patient testing, and clinical feedback.
The MVP would intentionally be a simple, useful personal health record that could involve some AI, but is not AI-heavy.
Patients could securely upload PDF health records—such as discharge summaries, physician notes, medication lists, laboratory reports, imaging reports, and specialist consultations—and organize them into a concise longitudinal record.
The application would present key information such as diagnoses, medications, allergies, major laboratory and imaging findings, procedures, hospitalizations, preventive-care history, and members of the care team in a format that is easy for patients to review and easy to share with clinicians.
The MVP could also generate a concise printable or shareable health summary so that, instead of bringing hundreds of pages to a new physician, a patient could provide the most important information in a few pages.
Using information already in the record—such as age, health conditions, vaccination history, and prior screening dates—the application would also identify common preventive services that may be due, including vaccinations, cancer screening, diabetes monitoring, cholesterol or blood-pressure monitoring, and other age- or risk-appropriate services based on established guidelines.
Some automated extraction or AI assistance may be useful, but the core MVP should not depend on sophisticated AI. Conventional software, structured forms, document parsing, and user confirmation could accomplish much of the essential functionality.
Starfield would serve as a real-world testing environment throughout the semester. The team could work with clinicians, staff, and patients to identify common use cases, test interface concepts, and see where users become confused.
The objective is not merely to build an application that works technically, but one that real patients can actually use.
Key questions include: Can someone with limited health literacy understand the screen without assistance? Can a patient who rarely uses technology upload a document? Can they easily see their medications, diagnoses, and preventive-care needs? Can they generate a useful health summary for a new physician in only a few taps?
AI would become much more important in later versions. A major stretch goal would be allowing patients to ask natural-language questions about their own medical history, such as: "What medications have I tried for my blood pressure?", "When was my last colonoscopy?", "What did my last CT scan show?", or "What should I tell my new cardiologist about my heart history?"
The application could search across years of records, provide a plain-language answer, and link statements back to the underlying source documents. Other AI-enabled goals could include specialty-specific summaries, explaining medical terminology, summarizing trends over time, identifying discrepancies between records, and helping patients prepare questions for appointments.
Longer-term features could include direct EHR connections, automated record import, medication and appointment tracking, laboratory and vital-sign trends, care-team contacts, and secure sharing of selected records. The team could also explore privacy-preserving architectures such as Torch’s mobile MCP model. These features would require additional clinical, legal, privacy, and security review.
We envision Pocket Doc becoming a widely used nonprofit health tool that helps low-income and vulnerable patients throughout the United States maintain control of their medical information as they move through an increasingly fragmented health care system.
The health care system often expects patients to serve as the connection between institutions that do not communicate well with one another. The patients least equipped to play that role are often the ones with the most complicated histories.
Their health history, in their pocket, organized around them.
Samuel S. Wang, Professor of Molecular Biology and the Princeton Neuroscience Institute
See a description of the Princeton Gerrymandering Project.
Stephen Kim, Associate Director for Information and Technology
The Princeton University Art Museum offers a world-class collection of over 100,000 works of art spanning the world of art from antiquity to the present. While more than 200,000 visitors visit our galleries in a year, we are always eager to develop new ways to engage audiences, especially, YOU, our students. Recently, we've built out new data and images services to power potential innovations like:
Samuel S. Wang, Professor of Molecular Biology and the Princeton Neuroscience Institute
Every 10 years, legislative districts across America must be redrawn after the Census. Redistricters have the task of making sure that diverse communities within a state are fairly represented. But they do not always know where those communities are.
Citizens have opportunities to testify about their communities in public hearings. But that testimony is qualitative, and there is no way to integrate the comments in a unified way. It would be useful to have a graphical application for individuals to (a) draw their communities of interest (COI's) on a state map, (b) store the shapes in a standard format such as GIS, and (c) annotate the shapes with comments. Then, after citizens have participated, it would be useful to display all of the communities of interest in a single map for inspection.
An additional feature might be reduction of redundancy by combining highly overlapping communities in a single consensus graphical display object.
Abby Klionsky '14, Office of the Executive Vice President
The decor in Frist -- all the quotes painted on the wall, etc. -- is meant to represent a diversity of ideas, and is one of the places on campus that, theoretically, does this quite well. It's theoretical because we don't know how much people actually pay attention to them, nor whether they know anything about the person being quoted.
There is actually documentation of all of this, in a very old-school, circa-2000 website that pairs photos of the quotes with photos and bios and explanations of the people who they are quoting: http://princeton.edu/frist/iconography.
This also covers the images in Cafe Viv and some of the Princeton-y flotsam that adorns the halls and walls. It would be GREAT if this could actually be a site that made people interested in looking at it!
Could we build a system that showed these images much more dynamically, perhaps with a rotating sequence of pictures that always showed something interesting. For each one, perhaps there could be a QR code that pointed to more details. Or maybe a touch screen would make it easy to get more details. Would it be possible to add new images and new text very easily without having to be an expert? Are there other things that would make the displays more appealing and encourage people to look at them more carefully?
Jill Stockwell, McGraw Center
Ideas that would greatly improve our organization's efficiency and communication. One is a volunteer application management system for our 150+ applicants each semester; another is a carpooling application for each of the seven facilities where we teach.
Wangyal Shawa, Map and Geospatial Information Center
We are planning two projects to create and manage our scanned maps and create geospatial data. One project is related to creating a batch georeferencing tool that will georeference scanned topographic maps that are the same size and the same scale. There is one system called QUAD-G (open source) to process the United States Geological Survey 1:24,000 scale maps but this software does not work well if you have a smaller scale map series. We need to customize the QUAD-G software to work with smaller scale maps using the same programming language or redesign it with a different programming language using similar workflows.
Another project is to design an open source software system that will extract georeferenced scanned maps to vector geospatial data.
These projects will benefit many researchers and libraries.
Ijeoma D. Nwagwu (ijeoma.nwagwu@princeton.edu), Office of Sustainability
The Office of Sustainability's Campus as Lab (CAL) program facilitates the use of Princeton's campus for sustainability research and experiential learning to advance the Sustainability Action Plan. Explorations into the social, physical, and operational dimensions of Princeton can generate new knowledge to help advance sustainability on campus, in our broader community, and around the world. Over the years COS 333 students have worked on several CAL projects and can support the Office of Sustainability on campus-based projects by developing:
Jed Marsh, Vice Provost for Institutional Research
There is an increasing interest in student outcomes after the initial placement -- say 10 years post degree. Currently, these data are harvested from a hodge-podge of sources, including scraping sites like LinkedIn. There's a fair amount of staff time spent across campus googling former students, both graduates and undergrads. We need tools that can:
Abby Klionsky '14, Office of the Executive Vice President
As a breakout group of the Campus Iconography Committee, the Princeton History Working Group is building a series of themed historical tours of Princeton's campus that will highlight lesser-known histories of the university. These will take shape in the form of a mobile app, which will use wayfinding technology to guide users to sites across campus and showcase associated photos, audio, and video to tell these stories. For some of these sites, we'd like to incorporate augmented reality features -- particularly in places where there may no longer be a physical marker or building still standing. The augmented reality component we're envisioning would likely be a statue for "placement" in one of the statue-hold pedestals in East Pyne courtyard or the front of Frist, a moving image to launch over a picture frame or screen that does exist in reality, or overlaying an old image of a campus map/building over what exists today.